Test Case Selection Based on a Spectrum of Complexity Metrics

Abstract

Quality continues to be a significant issue for software, and testing is a key activity in ensuring the quality of software. Unfortunately, since testing usually was near the end of software development, it is often rushed and frequently not done well. Thus, this paper suggests an approach that can address some of these issues and examine ways to reduce the software testing costs by selecting test cases based on a Spectrum of Complexity Metrics (SCM). We have developed a comprehensive taxonomy of product metrics based on two dimensions, the product level (class, method, statement) and the characteristics of the product. To evaluate these metrics, we have developed a tool which uses these metrics to target test cases. This tool enables us to identify three sample metric combinations. We have conducted a series of experiments based on three applications. The combinations of metrics were applied to test case selection on each of the applications. To investigate the efficiency of our test case selection, we have created a significant number of mutants using MuJava, and a series of a significant number of seeded errors inserted independently by a third party. Our experiments show that our test case selections discover at least 60 % of seeded errors and mutants. For further evaluation, we compared our approach to the boundary value analysis. Our experiments indicate that our testing approach is highly effective in the detection of the mutants as well as the seeded errors.